You are designing Azure Batch for a rendering workload. Job sizes vary widely, and you want to minimize cost using preemptible capacity but maintain a small always-on baseline. Tasks have dependencies and must be orchestrated per job submission. What should you implement? Choose two.

The correct options address both cost optimization and task orchestration requirements. Creating a pool with a small number of dedicated nodes and a larger number of low-priority nodes, combined with an autoscale formula that adjusts $TargetLowPriorityNodes based on pending tasks, effectively minimizes cost by leveraging cheaper preemptible capacity while ensuring a baseline of dedicated nodes for critical tasks. Configuring a Job Manager task allows for dynamic submission of tasks with dependencies, ensuring the correct execution order for complex rendering jobs. Incorrect options: Using a VM scale set outside Batch is not the native or recommended way to manage Batch compute resources. Disabling autoscaling and pre-provisioning a fixed-size pool would lead to either underutilization or insufficient capacity, failing to optimize cost or adapt to varying job sizes. Using a start task to open ports and manually scheduling tasks over SSH bypasses Batch's orchestration capabilities and is inefficient for large-scale processing.